Transcoders: cloud transcoding is growing in popularity, but sometimes only a dedicated appliance will do.
That's where transcoding comes in. At its basic level, a transcoder captures the stream of one type of codec and then converts it to another type of codec.
Transcoders sit squarely in the middle of the signal acquisition and viewing continuum. If an encoder captures the original signal and packetizes it, and a player receives the packets and converts them back to the original signal, then a transcoder acts as the translator between the acquisition codec and the list of codecs that the player has at its disposal for playback.
What About Converters?
We are sometimes asked about devices that will change a video signal from one format to another. These are more often referred to as converters--they're tasked with changing DVI to HDMI, NTSC to PAL, interlaced content to progressive frames, or even analog VGA to HDMI--since they don't actually packetize the video into segments or chunks that will be delivered via IP.
That said, there are some converters that have built-in encoding or transcoding, mainly for real-time delivery within a local area network (LAN) for image magnification, command and control, or digital signage. Even those products fall outside the realm of wide-area networks (WANs) or even internet-based video streaming.
Unified Communications and Real-Time Transcoding
One area we covered quite a bit in recent issues of Streaming Media magazine is the homogenization of enterprise video and its broader-scoped sibling: unified communications. In both of these areas, transcoding plays a central role.
Consider for a moment the need to live-stream a CEO all-hands meeting. For those unable to attend, the traditional H.264 encoding and streaming would suffice. Yet, for those who may not be able to view the content on a network, an audio-only call may be established. Most audio systems use either G.279 or G.711 audio compression, while most audio streaming systems use other audio codecs.
The same is true for playback of audio- or video-on-demand files. A quick glance at design and engineering questions on Stack Overflow (go2sm.com/stackoverflow) shows that solutions built around audioconferencing codecs--even for fairly ubiquitous player options, like ffmpeg and its ffplay player component--require some level of transcoding, since ffplay recognizes G.711 but not G.729 streams.
Doing this level of real-time transcoding for audio codecs is much less processor-intensive than video transcoding, which means most audio transcoding is now handled in software on commodity processors (and even on some mobile phones or tablets, given the dual-core or quad-core processor offerings on today's consumer smartphones).
Video transcoding, on the other hand, often still requires specialized hardware and dedicated software.
What About Media Servers?
Several transcoding features have moved toward the network core, though, in part due to the additional mission-critical role that media servers have been tasked with.
For several years now, media servers have been able to perform simple transcoding, such as decoding an MPEG-2 Transport Stream (M2TS) and converting the contained MPEG-2 primary video stream to a more modern streaming codec like Advanced Video Coding (AVC, aka H.264). But recent trends toward adaptive bitrate (ABR) video streaming have also put the burden of downscaling a single high-resolution stream into several lower bitrates and frame sizes.
An example of this, which is officially considered transrating, might be a live H.264 stream sent to the media server at 1080p30 (1920x1080 resolution at 30 frames per second). The transcoding, whether at a media server or via a standalone transcoder, would then convert the stream to at least one 720p30 (1280x720 resolution at 30 fps) and perhaps one standard-definition (SD) stream at 480p30 (854x480 at 30 fps).
In addition, either the transcoder or media server can perform packaging or segmentation steps, whereby the video files are broken down into very small files that can be delivered by a standard HTTP server--which scales much better than a traditional streaming protocol such as RTP or RTMP, which require a persistent session between the media server and the end-user's player.
These software transcoding tools typically are either built into the operating system, are open source, or are bundled with a video editing tool.
Built-in transcoding options are often limited to a very specific set of options. Apple's integrated Quick-Time player, up until recently, had a wide set of options for saving content in various formats. Recent versions, though, only offer options for transcoding or resizing to meet the form factors and resolutions of other Apple devices such as the iPad or iPhone.
Open source transcoders have increased in both quality and performance, but the very nature of using open source software means that some assembly is required. Due to limitations around licensing, most tools like HandBrake or ffmpeg require a number of additional modules to be downloaded and installed in order to transcode to still-popular legacy formats like MPEG-2 or VC1.
Bundled transcoding tools are geared toward assisting in the job of converting a very high-quality master file--such as 4K UHD or even RED 5K content, edited on a nonlinear editing (NLE) system--down to a bitrate and resolution that can be delivered as VOD on an over-the-top (OTT) device such as a mobile phone or streaming set-top box. Examples of these include Compressor from the Apple Final Cut Pro X NLE bundle or Adobe Media Encoder from the Adobe Premiere Pro NLE bundle.
For production workflows in which the NLE machine needs to be actively engaged in editing, server-based transcoding has become a much more affordable alternative.
Essentially, the server-based transcoding tools are the same software as their less-powerful desktop counterparts, but they include options such as watch folders (to assist in automated workflows) as well as some form of quality control and scheduling or prioritization of rush jobs. Open-source server-based transcoding tools, riding on Linux operating systems, have found their place in large media environments, thanks in part to the robustness of x264 and ffmpeg.
A number of companies make transcoding appliances, especially for higher resolutions (such as 4K Ultra HD) or live streams (including IP-to-IP transcoding). Some of the appliances also allow a number of files from a playlist--such as video ads, primary content, and secondary content that wasn't encoded in the desired streaming format or protocol--to be stitched together in a single on-demand (VOD) asset.
Another reason to use a transcoding appliance is the sheer amount of processing power, which is critical for doing faster-than-real-time transcodes of VOD assets. In the earlier playlist example, an hour's worth of content on a desktop transcoding software tool could take 3-4 hours to convert. On an appliance, though, it should take 15-20 minutes. Converting in one-fourth the time is a huge gap, and it may be worth paying the extra dollars in order to get optimized hardware assistance from a transcoding appliance.
Even companies that have traditionally been involved in cable TV signal modulation, such as Blonder Tongue Laboratories, Inc., are now offering transcoders that not only convert from cable standards such as ASI and QAM to IP-based ATSC formats, but they can also package and output HLS or even MPEG-DASH.
Assuming that the production pipeline includes a fat pipe to the outside world (i.e., a robust internet connection) then cloud-based encoding is a viable option. The master file, sometimes referred to as a mezzanine file, is uploaded to a transcoding service, and the various versions of streaming-ready files are converted in a scalable cloud-based environment.
Some cloud-based transcoding options include a desktop application that will segment large video files into smaller segments--although these segments aren't to be confused with stream-ready segments like those created by a packager for Apple HTTP Live Streaming (HLS) or MPEG Dynamic Adaptive Streaming via HTTP (DASH)--to make the overall mezzanine file uploading process go more smoothly. Be aware, though, that some of these desktop helper applications also do a bit of transcoding or transrating behind the scenes, potentially impacting the quality of your mezzanine file.
As we continue to see new video codecs (e.g., H.265, AV1) and new resolutions (e.g., 4K UHD or even 8K), the use of transcoding applications and appliances will continue to be relevant.
Desktop open source and bundled software transcoding applications are good for small transcoding jobs, but to really scale up and automate a workflow, careful attention should be paid to server- or cloud-based transcoding solutions. Appliances--dedicated to the task of transcoding, transrating, and repackaging VOD and live streams--continue to meet mission-critical transcoding needs, especially when it comes to new resolutions, compression formats, or faster-than-real-time transcoding requirements.
By Tim Siglin
Tim Siglin is a streaming industry veteran and longtime contributing editor to Streaming Media magazine.
Comments? Email us at firstname.lastname@example.org, or check the masthead for other ways to contact us.
|Printer friendly Cite/link Email Feedback|
|Title Annotation:||buyers' guides|
|Date:||Mar 1, 2017|
|Previous Article:||Encoding: from pico to infinite, encoders come in all shapes and sizes.|
|Next Article:||Quality analysis: whether it's video encoding or quality of service, you need to make sure your viewers are getting the best experience possible.|